Implementation of Suggestive Annotation: A Deep Active Learning Framework for Biomedical Image Segmentation This includes multiple uncertainty functions for active learning loop and implementation for dropout to estimate uncertainty instead of boostrap method described in the paper.
The architecture used is from DCAN: Deep Contour-Aware Networks for Accurate Gland Segmentation
The dataset used is the gland dataset. Its is available at: https://www2.warwick.ac.uk/fac/sci/dcs/research/tia/glascontest/download/warwick_qu_dataset_released_2016_07_08.zip. Download and unzip. Point to the folder in the run script(run.sh).
Some output from the model:
| Input | Output | Ground truth |
|---|---|---|
 |
 |
 |
TODO: description of each file
Active learning part might not be working!
I suspect that the some of the metrics are not calculated right, since the score does not really reflect the image outputted by the network.
- Make sure active learner is working
- Fix metrics
- Implement NasNet cells in the network
Shout-out to Rasmus Hvingelby, @hvingelby